Nuclear Regulatory Commission Report NUREG-75/014

From section 1: In the quantitative system probability estimates performed in this study, component behavior data in the form of failure rates and repair times are required as inputs to the system models. Since the goal of this study is risk assessment, as opposed to reliability analysis, larger errors (e.g. order of magnitude type accuracy) can be tolerated in the quantified results. This has important implications on the treatment of available data. In standard reliability analysis, point values (i.e., "best-estimates") are generally used for both data and results in quantifying the system model. In risk assessment, since results accurate to about an order of magnitude are sufficient, data and results using random variable and probabilistic approaches, can be usefully employed. The base of applicable failure rate data is thus significantly broadened since data with large error spreads and uncertainties can now be utilized. The data and associated material that were assembled for use in this study and that are presented here are to be used in the random variable framework (which will be described). The data and the accompanying framework are deemed sufficient for the study's needs. Care must be taken, however, since this data may not be sufficiently detailed, or accurate enough for use in general quantitative reliability models